Escherichia coli ATCC 25922 was the bacterial model for all experiments. Five E. coli clinical isolates, including two belonging to the high-risk clone ST131, harboring different combinations of chromosomally- and/or plasmid-mediated quinolone resistance mechanisms with susceptible, LLQR and resistance phenotypes were selected (Table 1; Briales et al., 2012; López-Cerero et al., 2013; Rodríguez-Martínez et al., 2016; Machuca et al., 2021). The SOS response was suppressed by recA gene knockout resistant to kanamycin, using a modified version of the method described by Datsenko and Wanner (2000) and Machuca et al. (2021). Liquid or solid lysogeny broth (LB; Invitrogen™, Madrid, Spain) medium and Mueller–Hinton broth (MHB; Oxoid, Madrid, Spain) were used. Strains were grown at 37°C. The antimicrobials used for the various assays were ciprofloxacin and kanamycin (Sigma–Aldrich, Madrid, Spain).

The susceptibility of each bacterial strain was determined in triplicate for each bacterial strain using the broth microdilution method, according to EUCAST guidelines. ¹ Briefly, an inoculum of 5 × 10⁵ CFU/mL of bacteria diluted in Mueller Hinton-Broth was exposed to twofold dilutions of ciprofloxacin. EUCAST 2023 (v13.0) clinical breakpoints were used for interpretation. Any change in MIC of at least two dilutions was considered significant. Clinical categories were established according to EUCAST breakpoints (Table 1; Machuca et al., 2021).

Whole-genome sequencing was performed to analyze the genomes of the 5 clinical isolates selected (FI 4, FI 10, FI 19, FI 20, and FI 24). Genomic DNA was extracted and sequenced on the MiSeq platform (Illumina, San Diego, CA, United States), and the library was prepared using the Nextera XT DNA library preparation kit (Illumina). Raw reads were quality filtered and assembled into contigs with CLC Genomics Workbench v.10.0 (Qiagen, Madrid, Spain). The average coverage was 50x. The resulting contigs were annotated with Prokka v. 1.14.5 (Seemann, 2014) using known proteins of E. coli ATCC 25922 from the UniProt release 2020_02 (The UniProt Consortium, 2018) as a reference database (“-proteins”), without removing the original annotation in case of conserved hypothetical proteins (“-rawproduct”), formatted according to NCBI standards (“-compliant -addgenes”) and annotating ncRNA elements using Rfam v. 14.1 (Kalvari et al., 2021; “-rfam”). ResFinder v.4.1 (Camacho et al., 2009; Zankari et al., 2017; Bortolaia et al., 2020) and MLST v.2. tools (Lemee et al., 2004; Bartual et al., 2005; Griffiths et al., 2010; Larsen et al., 2012) ² were used to identify acquired resistance genes [using an identity threshold of 90% (Zankari et al., 2012)] and determine the sequence type of each isolate, respectively. The genome assemblies were analyzed with OrthoFinder v.2.5.2 software (Emms and Kelly, 2019) to classify gene sequences into conserved gene families. Proteomic data obtained from the annotated genomes of the clinical isolates and reference strain were used as input data for OrthoFinder to find all clusters of orthologous groups (“-M msa -oa”; Emms and Kelly, 2015, 2019). All strain sequences were deposited in a public repository under accession number PRJNA1015411.

Escherichia coli strain ATCC 25922 and clinical isolates were grown at 37°C in LB medium to an OD600 of 0.5 (exponential phase). Cultures were diluted 10-fold in fresh LB. Ciprofloxacin was added to tubes at a final concentration of 1xMIC (relative to the wild-type MIC for each isogenic pair; Table 1). This concentration was sufficient to induce the relevant stress conditions without high lethality (Machuca et al., 2021) and allowed us at the same time to compare the cellular response to ciprofloxacin at identical absolute concentrations for each isogenic pair. All tubes were incubated at 37°C for 1 h with shaking. The remaining ciprofloxacin was removed by centrifugation at 10,000 × g for 2 min. After removal of the supernatant, an equivalent amount of fresh LB was added. Each experimental condition consisted of six independent replicates. One milliliter per sample was pelleted by centrifugation at 10.000 x g for 2 min. Cells were resuspended in 50 μL of TE (10 mM Tris–HCl pH 8.0, 1 mM EDTA) containing chicken lysozyme (0.1 mg/mL, Sigma Aldrich, Germany) and incubated for 5 min at room temperature with occasional swirling. A 250 μL volume of denaturation buffer (6 M urea/2 M thiourea in 10 mM HEPES pH 8.0) was added to each sample, and 25 μL (corresponding to approximately 50 μg of total protein) of the resulting lysate was used for in-solution protein digestion, as described previously (Rappsilber et al., 2007). Briefly, the proteins were re-suspended in denaturation buffer and reduced by the addition of 1 μL of 10 mM DTT dissolved in 50 mM ammonium bicarbonate (ABC) and incubated for 30 min, followed by a 20-min alkylation reaction in the dark by the addition of 1 μL of iodoacetamide at a stock concentration of 55 mM. As a first digestion step, 0.5 μg of Lysyl endopeptidase (LysC, Wako, Japan), resuspended in 50 mM ABC, was added and incubated for 3 h. After pre-digestion with LysC, the protein samples were diluted by a factor of 4 with 50 mM ABC (to reduce the concentration of urea) and subjected to overnight trypsin digestion at room temperature using 1 μg of sequencing grade modified trypsin (Promega, United States), also diluted in 50 mM ABC. The digestion was stopped by acidification by adding 5% acetonitrile and 0.3% trifluoroacetic acid (final concentrations). Samples were micro-purified and concentrated using the Stage-tip protocol, described elsewhere (Rappsilber et al., 2007), and eluates were dried under vacuum.

Peptides were reconstituted in 20 μL of 0.05% trifluoroacetic acid (TFA), 4% acetonitrile, and 5 μL were analyzed by an Ultimate 3,000 reversed-phase capillary nano liquid chromatography system connected to a Q Exactive HF mass spectrometer (Thermo Fisher Scientific). Samples were injected and concentrated on a trap column (PepMap100 C18, 3 μm, 100 Å, 75 μm i.d. × 2 cm, Thermo Fisher Scientific) equilibrated with 0.05% TFA in water. After switching the trap column inline, LC separations were performed on a capillary column (Acclaim PepMap100 C18, 2 μm, 100 Å, 75 μm i.d. × 25 cm, Thermo Fisher Scientific) at an eluent flow rate of 300 nL/min. Mobile phase A contained 0.1% formic acid in water, and mobile phase B contained 0.1% formic acid in 80% acetonitrile/20% water. The column was pre-equilibrated with 5% mobile phase B followed by an increase of 5%–44% mobile phase B in 100 min. Mass spectra were acquired in a data-dependent mode using a single MS survey scan (m/z 350–1,650) with a resolution of 60,000, and MS/MS scans of the 15 most intense precursor ions with a resolution of 15,000. The dynamic exclusion time was set to 20 s and automatic gain control was set to 3 × 10⁶ and 1 × 10⁵ for MS and MS/MS scans, respectively.

MS and MS/MS raw data were analyzed using the MaxQuant software package (version 2.0.3.0) with implemented Andromeda peptide search engine (Tyanova et al., 2016a). Data of the samples from strain ATCC 25922 were searched against the E. coli reference proteome downloaded from Uniprot (4,857 proteins, taxonomy 83,333, last modified 1 December 2019), while data of the samples from the 5 clinical isolates (FI 4, FI 10, FI 19, FI 20, and FI 24) were searched against individual databases generated from whole-genome sequencing as described above. The default parameters were used for MaxQuant except for enabling the options label-free quantification (LFQ) and match between runs. Filtering and statistical analysis was carried out for each strain individually using the software Perseus version 1.6.14 (Tyanova et al., 2016b). First, contaminants, reverse hits and ‘proteins only identified by site’ were removed from the dataset and protein LFQ intensities were log2 transformed. Next, two experimental groups (wild-type and recA mutant) were defined. Only proteins which were identified and quantified with LFQ intensity values in at least 3 (out of 6) experimental replicates (in at least 1/2 experimental groups) were included for downstream analysis. Missing protein intensity values were replaced from normal distribution (imputation) using the default settings in Perseus (width 0.3, down shift 1.8). Mean log₂ fold protein LFQ intensity differences between experimental groups (recA mutant—wild-type) were calculated in Perseus using a student’s t-test with permutation-based FDR of 0.05 to generate the adjusted p-values (=q-values). Proteins with a minimum 2-fold change in their relative intensity (log2-fold change > 1 for recA or log2-fold change < −1 for wild-type) and a q-value < 0.05 were considered significantly changed. Heatmaps and volcano plots were used to represent the results, using GraphPad Prism 8 software (Boston, Massachusetts United States).³

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository (Perez-Riverol et al., 2022)⁴ with the dataset identifier PXD050358.

The phenotypic and genetic characteristics of the isolates are shown in Table 1 (Machuca et al., 2021).

In total, 5,569 genes were found among the five selected clinical isolates and reference strain ATCC 25922: 3376 of these were present in all genomes, and 3,311 genes in a single copy. The total number of genes encoded by each isolate was 4,834, 4,811, 4,683, 4,701, 4,853, and 4,842 for ATCC, FI 4, FI 10, FI 19, FI 20, and FI 24, respectively. Between each isolate and ATCC 25922, the total number of genes in common was 4,482 (93.2%), 3,705 (79.1%), 3,647 (77.6%), 4,035 (83.1%), 4,043 (83.5%) for FI 4, FI 10, FI 19, FI 20, and FI 24, respectively.

Known and potential SOS-regulated genes described previously by Fernández De Henestrosa et al. (2000) and Courcelle et al. (2001) were found in the genomes of the selected isolates. Twenty-six of the 32 genes known to be LexA-regulated genes were present in all isolates, including ATCC 25922: umuC, umuD, sbmC, recN, urvB, dinI, recA, sulA, uvrA, uvrB, ssb, yebG, lexA, dinF, ydjQ, ruvA, ruvB, molR, uvrD, dinG, yigN, ydjM, ftsK, dinB, ybfE, polB. In addition, six of the 20 genes were identified as potential LexA-regulated genes: ymgF, ydeO, yoaA, yoaB, glvB, ibpA (Courcelle et al., 2001). Consequently, most of the SOS-regulated genes in E. coli were represented in our collection.

The protein expression of E. coli clinical isolates was compared with their isogenic pairs with suppressed SOS response under ciprofloxacin treatment at concentrations of 1xMIC relative to wild-type for 1 h.

Significant changes in protein expression were found for 460/1,702 proteins (27%) in ATCC, 664/1,509 (44%) in FI 4, 831/1,637 (51%) in FI 10, 904/1,726 (52%) in FI 19, 1,210/1,786 (68%) in FI 20, and 1,227/1,706 (72%) in FI 24 (see Supplementary Tables 1–6). The number of proteins that exhibited at least a significant 2-fold increase or decrease of their relative abundance upon recA deletion are highlighted in Figure 1. The number of proteins decreased upon recA deletion are: 93 for ATCC, 89 for FI 4, 46 for FI 10, 12 for FI 19, 74 for FI 20, and 111 for FI 24. The number of proteins increased upon recA deletion are: 12 for ATCC, 20 for FI 4, 10 for FI 10, 4 for FI 19, 18 for FI 20, and 32 for FI 24. Proteins with significant expression changes were plotted for each isolate and compared to the ATCC 25922 control strain (Figure 2), showing a similarity ranging between 4 and 22%. Regarding significant protein expression after suppression of the SOS system, 10 (DinG, DinI, RecA, RecN, RuvA, RuvB, SbmC, UmuD, UvrA, YebG) out of 32 proteins known whose genes are regulated by LexA were underexpressed in at least one isolate, and no potential protein regulated by LexA was affected.

The relative protein intensity between the recA mutant and wild-type are shown in Figure 3 for ATCC 25922 and the five clinical isolates and proteins that exhibit a very strong change in their abundance upon recA deletion are labeled. Underexpressed proteins (log2 FC < −2.5) were: RecA (as control), YebG, DinI, OmpW, PsuG, Oxc, ArgF, Ag43, DmsA, YfdX (for strain ATCC); RecA, SpeF, DinI, TdcE, YebG, AphA, PriS, DmsA, PsuG, RuvB, MalM, MalK, OmpW, GlpA, TdcF, CysN, YdfZ, RuvA, UvrA, RecN, PepE (strain FI 4); RecA, YebG, DinI, RpsQ, RecN, RbsA, UvrA, UmuD (for strain FI 10); RecA (strain FI 19); RecA, CadA, PsuK, PsuG, UgpB, RpmH, DinG, YfeC, CysP, TnaA, CysI, GlpT, DinI, TcyP (strain FI 20) and RecA, PsuG, DinI, YebG, CspD, RecN, GarL, TreB, SbmC, Hha, GlgS, SrlB (strain FI 24). Overexpressed proteins (log2 FC > 2.5) were: SpeF, TufB (for the ATCC strain); IbpB (strain FI 10); Ag43 (strain FI 19), RpmH, Hpf (strain FI 24; Figure 3).

Protein expression under ciprofloxacin pressure was highly variable both among isolates and between isolates and the reference strain. No relationship was observed between protein expression and the quinolone resistance phenotype displayed by the different isolates. At the ciprofloxacin concentration used, underexpressed proteins were more numerous than overexpressed ones, and were mainly associated with processes of DNA repair (RecA, YebG, DinI, DinG, RuvA, RuvB, UvrA, RecN, UmuD, CspD, SbmC) and energy production and conversion (DmsA, TdcE, AphA, TdcF, CysI, CysN, CysP, TnaA). Overexpressed proteins were few and were involved in different cellular processes, such as amino acid metabolism (SpeF), translation (TufB, RpmH), protein refolding (IbpB) and biofilm formation (Ag43). Taken together, these results indicate that the treatment had a serious impact on cellular physiology.

All proteins that showed as significant increase or decrease in their relative expression level upon recA deletion (log2 FC >2 or < −2) were analyzed in depth and classified by their function⁴ (Figure 4). Protein expression variations after treatment with ciprofloxacin most affected the following cellular functions: 15/76 (19.7%) proteins were involved in processes of energy production and conversion; 11/76 (14.5%) affected carbohydrate metabolism and transport; 7/76 (9.2%) were involved in replication, recombination and repair, cell wall and outer membrane structure and biogenesis; 6/76 (7.9%) proteins were involved in amino acid metabolism and transport, and 5/76 (6.6%) in DNA repair.

The following cellular processes were most affected and involved changes in expression of different proteins: energy production and conversion (AphA, CysH, CysI, CysN, CysP, DmsA, FdnG, FocA, HybO, MetF, TdcE, TdcF, TnaA, YdhR, YdhV); carbohydrate metabolism and transport (GarL, GlpA, GlpT, GltB, MalE, MalK, MalM, MglB, RbsA, SrlB, TreB, UgpB); replication, recombination and repair (CspD, RecA, RecX, RuvA, RuvB, SbmC, UmuD); cell wall and outer membrane structure and biogenesis (AmpD, BcsE, GlgS, LdtA, MlaE, OmpW, YhjG); amino acid metabolism and transport (ArgF, CadA, Hha, PepE, SpeF, TcyP) and DNA repair (DinG, DinI, RecN, UvrA, YebG; Figure 4).

Once again, the data indicate that the cellular abundance of a large number of proteins decreases under ciprofloxacin pressure in the absence of a functional SOS response.